Abstract：To fully explore the effective temporal information contained in power load data and improve the accuracy of short term power load prediction, this article proposes a power load forecasting model based on an improved attention mechanism for the temporal convolutional network (TCN)-long short term memory (LSTM) network. Firstly, the temporal data is input into the TCN model to extract temporal features. Then, the extracted temporal features are combined with non temporal to be input into the LSTM model for training. Finally, Bayesian optimization method is used for hyperparameter optimization to get the best parameters in TCN-LSTM. An attention mechanism improved by multi-layer perceptron (MLP) is introduced to reduce the loss of historical information and strengthen the influence of important information, completing short term load forecasting. By comparing the predictive performance of various deep learning models, it is verified that the model proposed in this article has higher accuracy in short term power load forecasting.
刘伟, 王洪志. 基于改进注意力机制的时间卷积网络-长短期记忆网络短期电力负荷预测[J]. 电气技术, 2024, 25(10): 8-14.
LIU Wei, WANG Hongzhi. Short term power load forecasting based on temporal convolutional network-long short term memory and improved attention mechanism. Electrical Engineering, 2024, 25(10): 8-14.
2024, Vol. 25 
